On-Line Thuriès et al.
789
Near infrared spectroscopy for predicting quality indices in the
organic fertiliser industry
Thuriès, L.
a, b; Bastianelli, D.
c; Davrieux, F.
d; Bonnal, L.
cand
Oliver, R.
ba Phalippou-Frayssinet S.A., Organic Fertilisers, F81240 Rouairoux, France.
E-mail: thuries@cirad.fr
b CIRAD, Laboratoire Matière Organique Sols Tropicaux, UPR078, TA70/01, F34398
Montpellier Cedex 05, France
c CIRAD, Laboratoire d’Alimentation Animale, TA 30/A Baillarguet, F34398 Montpellier
Cedex 5, France
d CIRAD, TA 80/16, F34398 Montpellier Cedex 5, France
Keywords: organic fertiliser, tropical plants, plant residues, Van Soest fractions, quality
index, C to N ratio
Introduction
The carbon to nitrogen (C/N) ratio can provide information on the capacity of an organic input to be transformed into humus [1 - 3], but it is not sufficient in some cases [4]. The lignin/nitrogen (Lig/N) ratio is another parameter used when modelling the transformation of organic materials [5]. An alternative estimate of degradability is the percentage of organic matter which is potentially resistant to mineralization over a long period of time. This estimate of the potentially humified organic matter (PHOM), calculated from an established equation based on the chemical composition, is normalised [7].
Quality indicators to determine the total carbon (C), nitrogen (N) and lignin content are expensive and time-consuming.
In this study we attempted to predict the PHOM index, the lignin/N ratio and the C/N ratio directly by near infrared (NIR) spectroscopy as a measure of the potential degradability of organic waste.
Materials and methods
Organic materials
The raw materials originated from industrially pre-processed plant residues and tropical plant residues were used. The pre-processed plant residues consisted of wet and dry grape skins, de-oiled grape pips, coffee cake, de-fatted cocoa bean, cocoa skin, olive pulp, maize cob, barley straw, rice hulls, rapeseed cake and soybean cake which were collected from the largest organic fertilizer factory in France. The tropical plant residue samples were collected from the field in Brazil and Kenya, as parts of trees, shrubs, crops and cover crops. These residues contained several plant parts including total above ground material, roots, stems, twigs, pods, leaves and litter, all of which were potentially utilisable for composting. The samples were chosen in order to encompass a wide variability of plant material.
Sample preparation and reference analyses
Each sample was analyzed for moisture content by drying to constant weight in an oven at 105°C, organic matter (OM) content as the loss in weight following ignition at 525°C overnight and total nitrogen (TN) content by the Kjeldahl method. Due to the heterogeneity of the fresh material, the samples were dried at 40°C and ground to pass through a 1 mm sieve for a sequential analysis of fibre content [8]. Briefly, each ground sample was successively extracted for neutral detergent fibre (NDF), acid detergent fibre (ADF) and acid detergent lignin (ADL). At each extraction step, the products obtained were filtered, dried at 40°C, weighed, and one replicate dried at 105°C for
Thuriès et al. On-Line
790
determining residual moisture, before ignition at 525°C to determine the ash content. Each fraction was calculated on an “ash free” basis.
The following indices were calculated;
(i) PHOM
:
Calculated from OM and the Van Soest fractions,
according to the French standardAFNORXPU44-
162 [7],PHOM = (0.3221 SOL – 0.7155 HEM + 0.6717CEL + 1.8919 LIC) x OM
x
10-
2+0.0271MM
where SOL = soluble fraction; HEM = hemicellulose fraction
;
CEL = cellulose fraction; LIC = lignin + cutin fraction in percent of the organic matter and MM = mineral matter. (ii) Lig/N: Calculated as ADL/TN and(iii) Ce/N: In theabsence of a measure of total carbon(C) an estimate of the carbonfraction (Ce
)
maybecalculatedasOM/2 to derive anapproximate estimateofthecarbon
/
nitrogen ratio. Calculated as OM/(2 x TN).Sample scanning and data analysis
Each sample was scanned on a NIRS 6500 (Foss NIRSystems, Silver Spring, MD, USA) in duplicate (2 different cup fillings) in ring cups. Spectral data were collected every 2 nm between 400 and 2,498 nm. The spectra, which each consisted of 32 scans, were stored as log (1/R) using a ceramic standard reference spectrum, then were corrected with a standard normal variate and detrend (SNVD) 2,5,5 [9] (Win-ISI, Infrasoft International, Port Matilda, PA, USA) mathematical treatment. Visible wavelengths were discarded because they introduced instability in the models. Calibrations of the studied parameters were performed using a modified partial least square regression (MPLS) [10] (WIN-ISI, Infrasoft International, Port Matilda, PA, USA), The standard error of calibration (SEC), the coefficient of determination (R²), and the standard error of validation (SECV) were calculated. In order to minimize over-fitting of the equations cross-validation was used as an internal cross-validation during calibration development.
Results and discussion
Figure 1 shows the reference values for the PHOM, Lig/N and Ce/N indices versus their
corresponding estimated values. Despite some outliers, the agreement between the reference and predicted values was particularly good for PHOM, but weaker for Lig/N and Ce/N. All the
parameters varied widely (Table 1), because of the diversity of the materials. Raw materials with a PHOM value under 25 can be used directed as an organic fertilizer, materials with a PHOM exceeding 40 have the potential to be transformed into humus in particular conditions. Plant residues with Ce/N values under 20 can be considered an organic fertilizer with a TN value exceeding 3. The
ideal Ce/N ratio for a mixture of composting materials in its early stages is about 25.
The models developed for the three parameters were accurate for PHOM and Lig/N, as their determination coefficients were above 0.92 and the ratio of their standard deviation over their SECV (ratio of performance to deviation (RPD)) were more than 3. The estimates were less accurate for Ce/N.
Furthermore, the good quality of the PHOM model can be partly explained by the PHOM reference values systematically covering a wide range, whereas there were gaps in the range of the Lig/N values. In addition the lower accuracy of the models developed for Lig/N and Ce/N could also be
partly explained by the fact that Ce is an estimate, although calculated as half of the organic matter
On-Line Thuriès et al. 791 Tr 0 20 40 60 80 100 120 0 20 40 60 80 100 120 SECV=5.96 Lig/N 0 10 20 30 40 0 10 20 30 40 SECV=2.36 Ce/N 0 10 20 30 40 0 10 20 30 40 SECV=1.94 Tr 0 20 40 60 80 100 120 0 20 40 60 80 100 120 SECV=5.96 Lig/N 0 10 20 30 40 0 10 20 30 40 SECV=2.36 Ce/N 0 10 20 30 40 0 10 20 30 40 SECV=1.94
Figure 1. Predictions derived using NIR spectroscopy of (a) PHOM (an index of the resistance of
carbon to mineralization over a long period of time), (b) the Lig/N ratio, and (c) the Ce/N ratio, for raw
materials used in the organic fertiliser industry derived from plant residues.
Table 1. Performance of PHOM, Lig/N and Ce/N calibration models for raw materials used in the organic fertiliser industry derived from plant residues.
Population Calibration Compost quality index
n Mean SD SEC R2 SECV RPD
PHOM 118 65.9 28.0 5.59 0.96 5.96 4.7
Lignin/Total N 76 16.2 7.2 1.83 0.94 2.36 3.1
Ce/Total N 164 23.4 4.9 1.71 0.88 1.94 2.5
n: number of samples SD: standard deviation
SEC: standard error of calibration
R2: coefficient of determination of calibration
SECV: standard error of cross-validation
RPD: ratio of performance to deviation (SD.SECV-1)
The accuracy of the model developed for PHOM is logical, as PHOM strongly depends on lignin. As a consequence this fraction is particularly well predicted by NIR spectroscopy as reported elsewhere in these Proceedings [11].
The differences between SEC and SECV (Table 1) were around 6%, 29% and 13% in relative value for PHOM, Lig/N and Ce/N, respectively. This tends to indicate that the model developed for
PHOM was the most robust. Even with a higher RPD, the model developed for Lig/N was less robust than the model for predicting Ce/N.
The SECV value for PHOM was above the within laboratory repeatability (SErep = 3.1) cited in the
French standard AFNOR XP U44-162 [7] for estimating PHOM. However, it was very close to the between laboratory repeatability (SErepro = 5.52) [7]. Furthermore, as the dry matter of the materials
can vary from 40 to 85%, the error would be between 2.38 and 5.06 g 100g-1 bulk weight, which is
very low in practice.
Considering each test costs approximately 200 Euros and takes approximately a week to obtain a PHOM result the precision of the NIR prediction is adequate for an industrial application.
Conclusion
It can be concluded that the PHOM and Lig/N indexes, used as quality estimators of compost and organic fertilisers, can be usefully estimated by NIR spectroscopy. The Ce/N index should rather be
used as a classification tool in the absence of true total C reference values. This model will be particularly useful for the organic fertiliser industry, as the mention of this index will be required to be stated on each product label in future (French Standard NFU 44-051 [12], under revision in 2006).
Thuriès et al. On-Line
792
References
1. E. J. Rubins and F. E. Bear, Soil Sci. 54, 411 (1942).
2. I. Trinsoutrot , S. Recous, B. Bentz, M. Linères, D. Chèneby and B. Nicolardot, Soil Sci.
Soc. Am. J. 64, 918 (2000).
3. M. Pansu, L. Thuriès, M.-C. Larré-Larrouy and P. Bottner, Soil Biol. Biochem. 35, 353 (2003).
4. S. Recous, C. Aita and B. Mary, Soil Biol. Biochem. 31, 119 (1999). 5. J. M. Melillo, J. D. Aber and J. F. Muratore, Ecology 63, 621 (1982). 6. D. Robin, Agronomie 17, 157 (1997).
7. XP U44-162 Norme expérimentale stabilité Amendements organiques, Association
Francaiçaise de Normalisation (AFNOR), Paris, France, PDF file (2004). 8. P. J. Van Soest, J. B. Robertson and B. A. Lewis, J. Dairy Sci. 74, 3583 (1991). 9. R. J. Barnes, M. S. Dhanoa and S. J. Lister, Appl. Spectrosc. 43, 772 (1989). 10. J. S. Shenk and M. O. Westerhaus, Crop Sci. 31, 469 (1991).
11. L. Thuries, D. Bastianelli, F. Davrieux, L. Bonnal and R. Oliver, in Near infrared
spectroscopy: Proceedings of the 12th International Conference. Ed by G.R. Burling-Claridge, S.L. Holroyd and R.M.W. Sumner, New Zealand Near Infrared Spectroscopy Society Inc, Hamilton New Zealand, p.781 (2007).
12. NFU No 44051 Amendements organiques, Association Française de Normalisation (AFNOR), Paris, France, pp. 694 (1981).